In RamiKrispin/forecastLM: Time Series Forecasting with Linear Regression Models

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width=8
)

The forecastLM package provides a framework for forecasting regular time series data with a linear regression model. This vignette introduces the basic forecasting process with the package. The following vignettes cover advanced functionalities of the package.

Basic forecasting

library(forecastLM)

data("ny_gas")

head(ny_gas)

class(ny_gas)

library(TSstudio)

ts_plot(ny_gas,
        title = "The New York Natural Gas Residential Monthly Consumption",
        Ytitle = "Million Cubic Feet",
        Xtitle = "Source: US Energy Information Administration (Jan 2020)")

md1 <- trainLM(input = ny_gas, 
              y = "y",
              seasonal = "month",
              trend = list(linear = TRUE))

names(md1)

summary(md1$model)

plot_fit(md1)

plot_res(md1)

events <- list(outlier = c(as.Date("2015-01-01"), as.Date("2015-02-01"), as.Date("2018-01-01"), as.Date("2019-01-01")))

md2 <- trainLM(input = ny_gas, 
              y = "y",
              seasonal = "month",
              trend = list(linear = TRUE),
              events = events)

plot_res(md2)

md3 <- trainLM(input = ny_gas, 
              y = "y",
              seasonal = "month",
              trend = list(linear = TRUE),
              events = events,
              lags = c(1,12))

plot_res(md3)

fc3 <- forecastLM(md3, h = 60)

plot_fc(fc3)

RamiKrispin/forecastLM documentation built on April 4, 2020, 1:48 a.m.